Radar Teach and Repeat: Architecture and Initial Field Testing

Abstract

Frequency-modulated continuous-wave (FMCW) scanning radar has emerged as an alternative to spinning LiDAR for state estimation on mobile robots. Radar's longer wavelength is less affected by small particulates, providing operational advantages in challenging environments such as dust, smoke, and fog. This paper presents Radar Teach and Repeat (RT&R): a full-stack radar system for long-term off-road robot autonomy. RT&R can drive routes reliably in off-road cluttered areas without any GPS. We benchmark the radar system's closed-loop path-tracking performance and compare it to its 3D LiDAR counterpart. 11.8 km of autonomous driving was completed without interventions using only radar and gyro for navigation. RT&R was evaluated on different routes with progressively less structured scene geometry. RT&R achieved lateral path-tracking root mean squared errors (RMSE) of 5.6 cm, 7.5 cm, and 12.1 cm as the routes became more challenging. On the robot we used for testing, these RMSE values are less than half of the width of one tire (24 cm). These same routes have worst-case errors of 21.7 cm, 24.0 cm, and 43.8 cm. We conclude that radar is a viable alternative to LiDAR for long-term autonomy in challenging off-road scenarios. The implementation of RT&R is open-source and available at: https://github.com/utiasASRL/vtr3.

Figures (5)

• Figure 1: Top: Our Clearpath Warthog UGV is equipped with Navtech RAS3 radar, Ouster OS1-128 LiDAR, and NovAtel RTK GPS system with a stationary receiver. Bottom: Birds-eye-view of Warthog repeating a path in a clearing of the Woody Loop. The robot is repeating within the tracks, visible from the teach. The associated radar point cloud highlights that Radar Teach and Repeat can drive precisely even in geometrically challenging regions.
• Figure 2: A factor-graph representation of the two odometry updates. Dark blue triangles represent states containing radar scans, light blue triangles represent states containing gyro measurements, black dots represent cost terms in the optimization, and grey ellipses capture which states and cost terms are active in each optimization problem. a) The 4 Hz radar odometry pipeline uses continuous-time ICP, a constant velocity motion model, and a preintegrated yaw rate measurement term. The scan times are denoted with $t_{s_n}$ and states at which scans occur are shown as large, dark blue triangles. b) Between consecutive radar scans, the odometry is updated at 100 Hz using raw gyro yaw rate measurements. Gyro odometry updates occur between scan times $t_{s_n}$.
• Figure 3: Satellite view of the four routes used for evaluation.
• Figure 4: Example repeats of the Parking, Mars Dome, and Grassy Loops using the RT&R pipeline. The absolute lateral path-tracking error along these repeats is highlighted using colours ranging from green (small error) to red (large error).
• Figure 5: Distribution of RT&R and LT&R lateral path-tracking errors across all repeats of three routes.